Sync meeting on software testing (27‐07‐2023)

Jump to bottom

Caspar van Leeuwen edited this page Jul 28, 2023 · 1 revision

EESSI test suite sync meetings

Planning

every 2 weeks on Thursday at 14:00 CE(S)T
next meetings:
- Thu 13 July 14:00 => Kenneth/Sam is on summer vacation, OK for Lara/Caspar/Satish
- Thu 27 July 14:00 => Kenneth is on summer vacation, OK for Lara/Caspar/Satish/Sam(maybe)
- Wed 9 Aug 14:00
- Fri 25 Aug 10:00
- Wed 6 Sept 10:30
- Wed 20 Sept 14:00

Meeting notes

TensorFlow test was reviewed by Sam, Caspar needs to look at it
- https://github.com/EESSI/test-suite/pull/38 has been reviewed.
- https://github.com/EESSI/test-suite/pull/60 needs to be merged before this.
- TensorFlow has issues with thread binding, currently binding MPI processes gives decent performance.
- New feature request for a custom launcher for system with hyperthreads enabled.(https://github.com/EESSI/test-suite/issues/74)
Arrange access to Vega for Satish Lara and Xin (if she doesn't have it yet)
- DPAs have been signed and Satish,Xin (,Lara?) already have received their logins.
Snellius config PR ready to be merged
- Merged.
Modify ReFrame easyconfig to add hpctestlib to the PYTHONPATH
- Created a easyconfig for this and has been merged in Easybuild.
- Rebuilding process hit a snag.
Rebuild ReFrame in EESSI
- This has hit a snag.
Caspar added his Vega scripts as a PR.
CPU auto detect on graviton nodes does not work on citc, rebuild of image required to build Python again.
- The autodetect works now and the PR is ready to be reviewed.
GROMACS regular runs are done on vega, but multinode failing => Caspar will look into it
- They have magically succeeded.
Do not define scales for the OSU tests and hard code num_tasks and num_tasks_per_node with tags 1_node and 2_nodes.
Collective communications default SCALES is applicable.
Hortense: $SLURM_PARTITIONS environment variable overrides the command line option --partition, was "supposed to be" solved with prepare_cmds but what actually worked was force unloading the sticky slurm environment environment module.
```
'prepare_cmds': ['export SLURM_PARTITION=gpu_rome_a100_40', 'source /cvmfs/pilot.eessi-hpc.org/versions/2023.06/init/bash']
```
Question for Vaselios: Are prepare commands also executed on the submit host? question asked

Previous meetings